Thanks for your review and comment.
No, the change was not intentional. Actually I should have been more careful when I made the change. I was thinking the overall overhead to run the tests is very small even if all ops have a larger default scale.
I don't think they all need this change. The v1 patch is to change the BlockReportStats only.
While debugging the code, I've no idea how numOpsRequired is explained in BlockReportStats (perhaps a bad design). When addBlock() adds blocks in datanode, the TinyDatanode is not really updating the blocks list (perhaps a bug). Moreover, the blockReportList is not built using the blocks either, so the block report will be empty (I comment in
HDFS-7435). All the above makes the BlockReportStats dubious to me, though this patch seems independent.